AI Nuclear: Signed Megawatts vs. Announced Promises

Zach Martin

Welcome to Nuclear Capital on I R Hub Radio. I'm Zach Martin, and I'm here with Emily Martin. Today we're asking one question that I think matters more than any single headline this month. When a tech company says it's going nuclear, how much of that is real, signed, money on the table, and how much is a press release with good intentions? There's a number floating around mid October industry tracking that frames it nicely. For every one megawatt under a signed contract with an AI or data center buyer, there are roughly 3.8 megawatts that have been announced but are nonbinding.

Emily Martin

Right, and the raw figures are worth saying out loud. About 10.8 gigawatts sits under executed, binding contracts, things like firm power purchase agreements and retail energy supply contracts. Another 40.4 gigawatts sits in announced, nonbinding paper, meaning memorandums of understanding, letters of intent, joint development agreements, early stage options. That's where the 3.8 to 1 ratio comes from. So the headline total looks enormous, north of 51 gigawatts, but only about one in five of those megawatts has a counterparty legally on the hook.

Zach Martin

I love a simple analogy for this. Telling your friends you're going to run a marathon is announced. Paying the entry fee, buying the shoes, and following the training plan every morning is signed. Both feel good in the moment, but only one of them gets you across the finish line. In capital markets, the entry fee is a binding contract.

Emily Martin

And here's the structural reason that matters. In project finance, a signed contract with an investment grade buyer, think Google, Amazon, Meta, Microsoft, is something commercial banks and private debt markets can actually underwrite. A long contract from a creditworthy buyer lowers the risk, and lower risk generally means a lower weighted average cost of capital. The nonbinding paper can't secure debt financing today. It works mainly as a demand signal, a way of telling the market that buyers want this, rather than deployable project capital.

Zach Martin

Now, I want to be fair, because nonbinding doesn't mean fake. A signal can be valuable. It tells suppliers, regulators, and developers where demand is heading. But as an analyst of your own decisions, you have to know which bucket you're looking at. So let's go through what's actually signed, because the recent examples are instructive. Emily, start us with Google and Constellation.

Emily Martin

This was announced October 6th. It's a 20 year power purchase agreement between Google and Constellation Energy that will enable investments at 11 Constellation owned nuclear units across six sites in Illinois, Pennsylvania, and New Jersey. The investments increase thermal and electrical efficiency, which unlocks about 890 megawatts of additional firm power. Constellation describes that as more than 4.3 billion dollars of new investment by Constellation, and the first uprate is expected to be delivered by 2028. Utility Dive pointed out the 890 megawatts is the output of several small modular reactors, or one large new conventional reactor, but it's achieved at plants that already exist.

Zach Martin

That's the point I keep coming back to. An uprate is like getting more out of the team you already have through better training and better equipment, instead of building a whole new stadium. Turbines, steam generators, digital controls. The release says uprates deliver capacity without the multi year timelines or interconnection bottlenecks of greenfield construction. And notice who's paying. Constellation's chairman, Joe Dominguez, called it funded by private entities.

Emily Martin

There's a grid policy angle too. The companies called it a direct response to PJM's Bring Your Own Power proposal, and they tied it to the White House Ratepayer Protection Pledge, the idea that big power users bring new supply rather than pushing costs onto residential customers. Utility Dive noted PJM is short about 6.8 gigawatts of new capacity, and that it's struggling to find resources as data center demand rises. The agreement also includes load shaping and demand response, so some non critical consumption can be curtailed during grid stress.

Zach Martin

One detail I want to flag, because it affects how you read that 10.8 gigawatt signed number. The Google deal also includes a separate 15 year agreement covering 2,700 megawatts of Constellation's existing PJM fleet. That's not new capacity. It's revenue certainty that keeps existing plants economically viable. So when you hear signed gigawatts, that's my interpretation here, not a statistic from the sources, some of them are existing plants getting a long term customer and some are genuinely new megawatts. Worth asking which one you're looking at every time.

Emily Martin

That's a good distinction. The Amazon and Constellation Calvert Cliffs deal fits the same pattern. Announced September 30th, it's a 20 year agreement for 690 megawatts from the 1,790 megawatt Calvert Cliffs plant in Maryland, and that 690 includes about 190 megawatts of uprate expected online between 2030 and 2032. Constellation says it supports more than 3 billion dollars of investment and helps position the plant for relicensing for another 20 years. And one clarification from Constellation: Amazon withdrew its earlier on site data center proposal, so this isn't behind the meter. All the electricity continues flowing to the PJM grid.

Zach Martin

Which is a nice illustration of how the money actually works. The buyer isn't plugging into the reactor. The buyer is signing a long contract that gives the plant owner confidence to spend the capital. Contract first, capital second. That's discipline.

Emily Martin

Then there's Meta and Vistra, which also connects to something we discussed in an earlier episode on this same announced versus signed theme. Back then, the 4.2 billion dollar federal financing for Vistra was early reporting we treated as unconfirmed. Now it's confirmed. On October 6th, the Department of Energy's Office of Energy Dominance Financing, which used to be the Loan Programs Office, announced a conditional loan commitment of 4.2 billion dollars to Vistra for uprates and modernization at its nuclear plants in Pennsylvania and Ohio.

Zach Martin

And the contract underneath it is from January. Meta signed 20 year agreements with Vistra: the full 1,268 megawatts from Perry, the full 908 megawatts from Davis Besse, both in Ohio, plus uprate capacity of 213 megawatts at Perry, 80 at Davis Besse, and 140 at Beaver Valley in Pennsylvania. That's 433 megawatts of uprates, expected in the early 2030s. Add it up and the 2.6 gigawatt number checks out.

Emily Martin

One qualifier we should keep attached, and DOE was explicit about it: this is a conditional commitment. Vistra must satisfy technical, legal, environmental, and financial conditions before DOE enters definitive financing documents and funds the loan. DOE also said the loan would cover the 433 megawatts of new capacity and preserve around 4 gigawatts of existing nuclear, supporting operation for 20 years beyond the plants' existing licenses. There's also an option for future uprates at Comanche Peak in Texas. So the signed contract came first, and the federal capital is now lining up behind it, but it isn't fully funded yet.

Zach Martin

So the signed side of the ledger is real, but look at its character. It's mostly existing plants. Uprates. Single digit to a few hundred megawatts each. Practical, near term, bankable. That's the compounding approach, small improvements stacked over time. Now the other 40 gigawatts. That's where it gets more speculative, and more exciting.